# Help me addict myselfMay 30, 2012 11:00 AM   Subscribe

How do you determine the ideal variable reinforcement rate?

I know that to promote an activity, the best way is via variable reinforcement with a slow decrease in the rate of rewards. However, I can't figure out what a good starting rate is. I understand it would be determined by the activity, but if I was looking for a variable reinforcement rate for a daily activity, what would be a good rate?

(I'm basing this idea on obiwanwasabi's comment and am trying it with myself. I have started with a 1 in 50 chance if I spend an hour on the project I am working on but am unsure if this is so infrequent that it may lead to extinction behavior.)

I don't have easy access to the academic material, and have been told to read Don't Shoot the Dog, which I will get soon. What I wish I could see is whatever companies like Zynga or Blizzard use to determine the drop rates in their online games, but I assume that stuff is confidential.
posted by Hactar to Science & Nature (4 answers total) 3 users marked this as a favorite

It won't be a fixed rate. The reinforcements need to be *variable*, i.e. random, so that you can't predict when they're coming. I don't know what the mimimum frequency ought to be, but it probably depends on how much you like the reinforcement. Also, keep in mind that rewards and reinforcements are not the same thing. A reinforcement is relatively immediate, while a reward is delayed. I'm not sure variable reward schedules would be as effective as variable reinforcement schedules.
posted by jon1270 at 1:40 PM on May 30, 2012 [2 favorites]

I find that I like playing games where I have a lot of initial success and then they get progressively harder. So, maybe something more like 1/2t/10 — t being the number of times you’ve “spun the wheel.”

Of course it depends on the value of your reinforcement. Perhaps you could look at the expected value of the first iteration and see how you feel about it? For example, if your reinforcement is a cookie, is 1/50 of a cookie worth working an hour?

posted by danielparks at 4:56 PM on May 30, 2012

danielparks, the reward right now is Railsea by China Mieville. The project is actually learning python enough to automate the process.

I think I'm going to lower it down to 1 in 20 with a slightly lower value reward and scale up the rewards as I increase the odds. I'm thinking 1 in 30 for the second reward, etc.

And I won't even cheat by using my lucky d20 from D&D.
posted by Hactar at 8:39 AM on May 31, 2012

Give yourself a much smaller reward, much more frequently. A lolly, or something happy & silly, say 1/2, or 1/3 times. Possibly even doing a victory dance regularly. Anything that feels like, yeah! Success!

Basically, much, much more frequent reinforcement than 1 in 30 hours.
posted by Elysum at 4:02 AM on June 29, 2012

« Older Is it tacky to ask an ex-vendor for tickets?   |   Remove WD-40 smell from laundry. Newer »